Issue description

When dir(torch) fails with the following error message: TypeError: descriptor '__dir__' of 'object' object needs an argument. Since __dir__ underlies inspect, pydoc, etc., it means tab completion at the module level is broken in interactive environments.

Code example

>>> import torch
>>> dir(torch)
Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
TypeError: descriptor '__dir__' of 'object' object needs an argument

System Info

This was a fresh install (yesterday, 1 Aug 2018) of both Python3 (via homebrew) and pytorch (via pip) on a new laptop.

Collecting environment information...
PyTorch version: 0.4.1
Is debug build: No
CUDA used to build PyTorch: None

OS: Mac OSX 10.13.4
GCC version: Could not collect
CMake version: Could not collect

Python version: 3.7
Is CUDA available: No
CUDA runtime version: No CUDA
GPU models and configuration: No CUDA
Nvidia driver version: No CUDA
cuDNN version: No CUDA

Versions of relevant libraries:
[pip] Could not collect
[conda] Could not collect